tcg/i386: Adjust TCG_TARGET_HAS_MEMORY_BSWAP

Always true when movbe is available, otherwise leave
this to generic code.

Backports d2ef1b83a7a2047e0e36d7b62b3a5d151ab958f5

qemu/aarch64.h

@@ -884,6 +884,7 @@
 #define have_bmi1 have_bmi1_aarch64
 #define have_bmi2 have_bmi2_aarch64
 #define have_popcnt have_popcnt_aarch64
+#define have_movbe have_movbe_aarch64
 #define hcr_write hcr_write_aarch64
 #define helper_access_check_cp_reg helper_access_check_cp_reg_aarch64
 #define helper_add_saturate helper_add_saturate_aarch64

qemu/aarch64eb.h

@@ -884,6 +884,7 @@
 #define have_bmi1 have_bmi1_aarch64eb
 #define have_bmi2 have_bmi2_aarch64eb
 #define have_popcnt have_popcnt_aarch64eb
+#define have_movbe have_movbe_aarch64eb
 #define hcr_write hcr_write_aarch64eb
 #define helper_access_check_cp_reg helper_access_check_cp_reg_aarch64eb
 #define helper_add_saturate helper_add_saturate_aarch64eb

qemu/arm.h

@@ -884,6 +884,7 @@
 #define have_bmi1 have_bmi1_arm
 #define have_bmi2 have_bmi2_arm
 #define have_popcnt have_popcnt_arm
+#define have_movbe have_movbe_arm
 #define hcr_write hcr_write_arm
 #define helper_access_check_cp_reg helper_access_check_cp_reg_arm
 #define helper_add_saturate helper_add_saturate_arm

qemu/armeb.h

@@ -884,6 +884,7 @@
 #define have_bmi1 have_bmi1_armeb
 #define have_bmi2 have_bmi2_armeb
 #define have_popcnt have_popcnt_armeb
+#define have_movbe have_movbe_armeb
 #define hcr_write hcr_write_armeb
 #define helper_access_check_cp_reg helper_access_check_cp_reg_armeb
 #define helper_add_saturate helper_add_saturate_armeb

qemu/header_gen.py

@@ -890,6 +890,7 @@ symbols = (
     'have_bmi1',
     'have_bmi2',
     'have_popcnt',
+    'have_movbe',
     'hcr_write',
     'helper_access_check_cp_reg',
     'helper_add_saturate',

qemu/m68k.h

@@ -884,6 +884,7 @@
 #define have_bmi1 have_bmi1_m68k
 #define have_bmi2 have_bmi2_m68k
 #define have_popcnt have_popcnt_m68k
+#define have_movbe have_movbe_m68k
 #define hcr_write hcr_write_m68k
 #define helper_access_check_cp_reg helper_access_check_cp_reg_m68k
 #define helper_add_saturate helper_add_saturate_m68k

qemu/mips.h

@@ -884,6 +884,7 @@
 #define have_bmi1 have_bmi1_mips
 #define have_bmi2 have_bmi2_mips
 #define have_popcnt have_popcnt_mips
+#define have_movbe have_movbe_mips
 #define hcr_write hcr_write_mips
 #define helper_access_check_cp_reg helper_access_check_cp_reg_mips
 #define helper_add_saturate helper_add_saturate_mips

qemu/mips64.h

@@ -884,6 +884,7 @@
 #define have_bmi1 have_bmi1_mips64
 #define have_bmi2 have_bmi2_mips64
 #define have_popcnt have_popcnt_mips64
+#define have_movbe have_movbe_mips64
 #define hcr_write hcr_write_mips64
 #define helper_access_check_cp_reg helper_access_check_cp_reg_mips64
 #define helper_add_saturate helper_add_saturate_mips64

qemu/mips64el.h

@@ -884,6 +884,7 @@
 #define have_bmi1 have_bmi1_mips64el
 #define have_bmi2 have_bmi2_mips64el
 #define have_popcnt have_popcnt_mips64el
+#define have_movbe have_movbe_mips64el
 #define hcr_write hcr_write_mips64el
 #define helper_access_check_cp_reg helper_access_check_cp_reg_mips64el
 #define helper_add_saturate helper_add_saturate_mips64el

qemu/mipsel.h

@@ -884,6 +884,7 @@
 #define have_bmi1 have_bmi1_mipsel
 #define have_bmi2 have_bmi2_mipsel
 #define have_popcnt have_popcnt_mipsel
+#define have_movbe have_movbe_mipsel
 #define hcr_write hcr_write_mipsel
 #define helper_access_check_cp_reg helper_access_check_cp_reg_mipsel
 #define helper_add_saturate helper_add_saturate_mipsel

qemu/powerpc.h

@@ -884,6 +884,7 @@
 #define have_bmi1 have_bmi1_powerpc
 #define have_bmi2 have_bmi2_powerpc
 #define have_popcnt have_popcnt_powerpc
+#define have_movbe have_movbe_powerpc
 #define hcr_write hcr_write_powerpc
 #define helper_access_check_cp_reg helper_access_check_cp_reg_powerpc
 #define helper_add_saturate helper_add_saturate_powerpc

qemu/riscv32.h

@@ -884,6 +884,7 @@
 #define have_bmi1 have_bmi1_riscv32
 #define have_bmi2 have_bmi2_riscv32
 #define have_popcnt have_popcnt_riscv32
+#define have_movbe have_movbe_riscv32
 #define hcr_write hcr_write_riscv32
 #define helper_access_check_cp_reg helper_access_check_cp_reg_riscv32
 #define helper_add_saturate helper_add_saturate_riscv32

qemu/riscv64.h

@@ -884,6 +884,7 @@
 #define have_bmi1 have_bmi1_riscv64
 #define have_bmi2 have_bmi2_riscv64
 #define have_popcnt have_popcnt_riscv64
+#define have_movbe have_movbe_riscv64
 #define hcr_write hcr_write_riscv64
 #define helper_access_check_cp_reg helper_access_check_cp_reg_riscv64
 #define helper_add_saturate helper_add_saturate_riscv64

qemu/sparc.h

@@ -884,6 +884,7 @@
 #define have_bmi1 have_bmi1_sparc
 #define have_bmi2 have_bmi2_sparc
 #define have_popcnt have_popcnt_sparc
+#define have_movbe have_movbe_sparc
 #define hcr_write hcr_write_sparc
 #define helper_access_check_cp_reg helper_access_check_cp_reg_sparc
 #define helper_add_saturate helper_add_saturate_sparc

qemu/sparc64.h

@@ -884,6 +884,7 @@
 #define have_bmi1 have_bmi1_sparc64
 #define have_bmi2 have_bmi2_sparc64
 #define have_popcnt have_popcnt_sparc64
+#define have_movbe have_movbe_sparc64
 #define hcr_write hcr_write_sparc64
 #define helper_access_check_cp_reg helper_access_check_cp_reg_sparc64
 #define helper_add_saturate helper_add_saturate_sparc64

qemu/tcg/i386/tcg-target.h

@@ -101,6 +101,7 @@ extern bool have_bmi1;
 extern bool have_popcnt;
 extern bool have_avx1;
 extern bool have_avx2;
+extern bool have_movbe;
 
 // UNICORN FIXME:
 // Taken from cpuid.h in mainline qemu.
@@ -262,7 +263,7 @@ static inline void tb_target_set_jmp_target(uintptr_t tc_ptr,
 
 #define TCG_TARGET_DEFAULT_MO (TCG_MO_ALL & ~TCG_MO_ST_LD)
 
-#define TCG_TARGET_HAS_MEMORY_BSWAP  1
+#define TCG_TARGET_HAS_MEMORY_BSWAP  have_movbe
 
 #ifdef CONFIG_SOFTMMU
 #define TCG_TARGET_NEED_LDST_LABELS

qemu/tcg/i386/tcg-target.inc.c

@@ -163,13 +163,12 @@ bool have_bmi1;
 bool have_popcnt;
 bool have_avx1;
 bool have_avx2;
+bool have_movbe;
 
 #ifdef CONFIG_CPUID_H
-static bool have_movbe;
 static bool have_bmi2;
 static bool have_lzcnt;
 #else
-# define have_movbe 0
 # define have_bmi2 0
 # define have_lzcnt 0
 #endif
@@ -1954,13 +1953,14 @@ static void tcg_out_qemu_ld_direct(TCGContext *s, TCGReg datalo, TCGReg datahi,
                                    TCGReg base, int index, intptr_t ofs,
                                    int seg, bool is64, MemOp memop)
 {
-    const MemOp real_bswap = memop & MO_BSWAP;
+    bool use_movbe = false;
-    MemOp bswap = real_bswap;
     int rexw = is64 * P_REXW;
     int movop = OPC_MOVL_GvEv;
 
-    if (s->have_movbe && real_bswap) {
+    /* Do big-endian loads with movbe.  */
-        bswap = 0;
+    if (memop & MO_BSWAP) {
+        tcg_debug_assert(have_movbe);
+        use_movbe = true;
         movop = OPC_MOVBE_GyMy;
     }
 
@@ -1974,23 +1974,28 @@ static void tcg_out_qemu_ld_direct(TCGContext *s, TCGReg datalo, TCGReg datahi,
                                  base, index, 0, ofs);
         break;
     case MO_UW:
-        tcg_out_modrm_sib_offset(s, OPC_MOVZWL + seg, datalo,
+        if (use_movbe) {
-                                 base, index, 0, ofs);
+            /* There is no extending movbe; only low 16-bits are modified.  */
-        if (real_bswap) {
+            if (datalo != base && datalo != index) {
-            tcg_out_rolw_8(s, datalo);
+                /* XOR breaks dependency chains.  */
-        }
+                tgen_arithr(s, ARITH_XOR, datalo, datalo);
-        break;
-    case MO_SW:
-        if (real_bswap) {
-            if (s->have_movbe) {
                 tcg_out_modrm_sib_offset(s, OPC_MOVBE_GyMy + P_DATA16 + seg,
                                          datalo, base, index, 0, ofs);
             } else {
-                tcg_out_modrm_sib_offset(s, OPC_MOVZWL + seg, datalo,
+                tcg_out_modrm_sib_offset(s, OPC_MOVBE_GyMy + P_DATA16 + seg,
-                                         base, index, 0, ofs);
+                                         datalo, base, index, 0, ofs);
-                tcg_out_rolw_8(s, datalo);
+                tcg_out_ext16u(s, datalo, datalo);
             }
-            tcg_out_modrm(s, OPC_MOVSWL + rexw, datalo, datalo);
+        } else {
+            tcg_out_modrm_sib_offset(s, OPC_MOVZWL + seg, datalo,
+                                     base, index, 0, ofs);
+        }
+        break;
+    case MO_SW:
+        if (use_movbe) {
+            tcg_out_modrm_sib_offset(s, OPC_MOVBE_GyMy + P_DATA16 + seg,
+                                     datalo, base, index, 0, ofs);
+            tcg_out_ext16s(s, datalo, datalo, rexw);
         } else {
             tcg_out_modrm_sib_offset(s, OPC_MOVSWL + rexw + seg,
                                      datalo, base, index, 0, ofs);
@@ -1998,18 +2003,12 @@ static void tcg_out_qemu_ld_direct(TCGContext *s, TCGReg datalo, TCGReg datahi,
         break;
     case MO_UL:
         tcg_out_modrm_sib_offset(s, movop + seg, datalo, base, index, 0, ofs);
-        if (bswap) {
-            tcg_out_bswap32(s, datalo);
-        }
         break;
 #if TCG_TARGET_REG_BITS == 64
     case MO_SL:
-        if (real_bswap) {
+        if (use_movbe) {
-            tcg_out_modrm_sib_offset(s, movop + seg, datalo,
+            tcg_out_modrm_sib_offset(s, OPC_MOVBE_GyMy + seg, datalo,
                                      base, index, 0, ofs);
-            if (bswap) {
-                tcg_out_bswap32(s, datalo);
-            }
             tcg_out_ext32s(s, datalo, datalo);
         } else {
             tcg_out_modrm_sib_offset(s, OPC_MOVSLQ + seg, datalo,
@@ -2021,12 +2020,9 @@ static void tcg_out_qemu_ld_direct(TCGContext *s, TCGReg datalo, TCGReg datahi,
         if (TCG_TARGET_REG_BITS == 64) {
             tcg_out_modrm_sib_offset(s, movop + P_REXW + seg, datalo,
                                      base, index, 0, ofs);
-            if (bswap) {
-                tcg_out_bswap64(s, datalo);
-            }
         } else {
-            if (real_bswap) {
+            if (use_movbe) {
-                int t = datalo;
+                TCGReg t = datalo;
                 datalo = datahi;
                 datahi = t;
             }
@@ -2041,14 +2037,10 @@ static void tcg_out_qemu_ld_direct(TCGContext *s, TCGReg datalo, TCGReg datahi,
                 tcg_out_modrm_sib_offset(s, movop + seg, datalo,
                                          base, index, 0, ofs);
             }
-            if (bswap) {
-                tcg_out_bswap32(s, datalo);
-                tcg_out_bswap32(s, datahi);
-            }
         }
         break;
     default:
-        tcg_abort();
+        g_assert_not_reached();
     }
 }
 
@@ -2114,24 +2106,27 @@ static void tcg_out_qemu_st_direct(TCGContext *s, TCGReg datalo, TCGReg datahi,
                                    TCGReg base, int index, intptr_t ofs,
                                    int seg, MemOp memop)
 {
-    /* ??? Ideally we wouldn't need a scratch register.  For user-only,
-       we could perform the bswap twice to restore the original value
-       instead of moving to the scratch.  But as it is, the L constraint
-       means that TCG_REG_L0 is definitely free here.  */
     const TCGReg scratch = TCG_REG_L0;
-    const MemOp real_bswap = memop & MO_BSWAP;
+    bool use_movbe = false;
-    MemOp bswap = real_bswap;
     int movop = OPC_MOVL_EvGv;
 
-    if (s->have_movbe && real_bswap) {
+    /*
-        bswap = 0;
+     * Do big-endian stores with movbe or softmmu.
+     * User-only without movbe will have its swapping done generically.
+     */
+    if (memop & MO_BSWAP) {
+        tcg_debug_assert(have_movbe);
+        use_movbe = true;
         movop = OPC_MOVBE_MyGy;
     }
 
     switch (memop & MO_SIZE) {
     case MO_8:
-        /* In 32-bit mode, 8-bit stores can only happen from [abcd]x.
+        /*
-           Use the scratch register if necessary.  */
+         * In 32-bit mode, 8-bit stores can only happen from [abcd]x.
+         * TODO: Adjust constraints such that this is is forced,
+         * then we won't need a scratch at all for user-only.
+         */
         if (TCG_TARGET_REG_BITS == 32 && datalo >= 4) {
             tcg_out_mov(s, TCG_TYPE_I32, scratch, datalo);
             datalo = scratch;
@@ -2140,43 +2135,19 @@ static void tcg_out_qemu_st_direct(TCGContext *s, TCGReg datalo, TCGReg datahi,
                                  datalo, base, index, 0, ofs);
         break;
     case MO_16:
-        if (bswap) {
-            tcg_out_mov(s, TCG_TYPE_I32, scratch, datalo);
-            tcg_out_rolw_8(s, scratch);
-            datalo = scratch;
-        }
         tcg_out_modrm_sib_offset(s, movop + P_DATA16 + seg, datalo,
                                  base, index, 0, ofs);
         break;
     case MO_32:
-        if (bswap) {
-            tcg_out_mov(s, TCG_TYPE_I32, scratch, datalo);
-            tcg_out_bswap32(s, scratch);
-            datalo = scratch;
-        }
         tcg_out_modrm_sib_offset(s, movop + seg, datalo, base, index, 0, ofs);
         break;
     case MO_64:
         if (TCG_TARGET_REG_BITS == 64) {
-            if (bswap) {
-                tcg_out_mov(s, TCG_TYPE_I64, scratch, datalo);
-                tcg_out_bswap64(s, scratch);
-                datalo = scratch;
-            }
             tcg_out_modrm_sib_offset(s, movop + P_REXW + seg, datalo,
                                      base, index, 0, ofs);
-        } else if (bswap) {
-            tcg_out_mov(s, TCG_TYPE_I32, scratch, datahi);
-            tcg_out_bswap32(s, scratch);
-            tcg_out_modrm_sib_offset(s, OPC_MOVL_EvGv + seg, scratch,
-                                     base, index, 0, ofs);
-            tcg_out_mov(s, TCG_TYPE_I32, scratch, datalo);
-            tcg_out_bswap32(s, scratch);
-            tcg_out_modrm_sib_offset(s, OPC_MOVL_EvGv + seg, scratch,
-                                     base, index, 0, ofs + 4);
         } else {
-            if (real_bswap) {
+            if (use_movbe) {
-                int t = datalo;
+                TCGReg t = datalo;
                 datalo = datahi;
                 datahi = t;
             }
@@ -2187,7 +2158,7 @@ static void tcg_out_qemu_st_direct(TCGContext *s, TCGReg datalo, TCGReg datahi,
         }
         break;
     default:
-        tcg_abort();
+        g_assert_not_reached();
     }
 }
 

qemu/x86_64.h

@@ -884,6 +884,7 @@
 #define have_bmi1 have_bmi1_x86_64
 #define have_bmi2 have_bmi2_x86_64
 #define have_popcnt have_popcnt_x86_64
+#define have_movbe have_movbe_x86_64
 #define hcr_write hcr_write_x86_64
 #define helper_access_check_cp_reg helper_access_check_cp_reg_x86_64
 #define helper_add_saturate helper_add_saturate_x86_64